Overall, the analyzed cohort was comprised of mostly WHO grade IV patients. In most of the studies the entire cohort received adjunctive chemoradiation at the time of initial resection. Five studies did not report 100% adjuvant chemotherapy rate, with one of them published prior to 2005 and the adoption of the Stupp protocol, and the other four not providing any further details. However, all patients included in these studies received EBRT following initial resection. At the time of reccurence the median prescription dose across the studies was surprisingly low. Few studies reported any grade ≥ 3 radiation toxicity or radiation necrosis. At one year of clinical follow slightly more than half of the patients in the pooled analysis were alive. Consistent with previous literature local control was more frequent than distant control, however distant control was demonstrated in greater than a quarter of patients in the pooled analysis. This highlights the importance distant control in rHGGs, emphasizing the role of systemic therapy in this setting. Unfortunately, data on the chemotherapy protocols was limited across the studies.

Identifying prognostic factors and intervention protocols to prolong survival in patients with rHGG is a persistent challenge. The current estimate of overall median survival following SRS for rHGG is 10.6 months as demonstrated in our analysis. Comparatively surgery and chemotherapy for rHGG demonstrate overall survival of 3–13 months and 5–6 months respectively [10,11,12,13,14,15,16,17,18,19,20]. A frequently observed predictor of improved survival in patients with rHGG undergoing SRS is the interval between surgery for the primary tumor reccurence [4, 21]. This finding is perhaps a reflection of more indolent biology. Imber et al. defined an interval of 20.2 months from the time of surgery to SRS as significantly associated with improved survival [22]. A more conservative estimate was reported by Lovo et al. with a survival benefit for patients undergoing SRS ≥ 10 months from surgery [21]. Given that most recurrences occur prior to 20 months the findings by this study are more applicable. Performance status as measured by the Karnofsky Performance Status (KPS) consistently predicts survival in this patient population. However, the threshold at which KPS predicts improved survival is variable with reports ranging from KPS > 70 to KPS > 90 [4]. Age at the time of SRS is also reported to be associated with survival outcomes [2, 15]. Increasing age, which may be related to declining performance status is invariably associated with diminished survival. Tumors designated as WHO grade III have a relatively prolonged survival compared to grade IV tumors, which may be attributable to more aggressive growth dynamics of grade 4 tumors [23]. Finally, tumor volume at the time of SRS may predict survival with increased tumor volume at the time of SRS associated with diminished survival. Overall, younger age, higher KPS, and smaller tumor volume are persistently associated with improved survival outcomes. Age demonstrated a strong relationship as studies with ≥ 70% survival at 1-year reported a median age of < 60 years across each of these studies.

The use of single fraction over hypofractionated SRS, volumetric constraints, dose escalation, and tumor location have been described as primary factors influencing response to SRS in rHGG [15, 24,25,26]. Presently there is no clear evidence supporting the use of a single session fraction SRS versus fractionated SRS (fSRS). Fractionation has theoretical radiobiological advantages over single session SRS. Administering SRS over multiple sessions allows for larger tumor volumes to be treated with higher radiation doses [15, 24, 27]. Also, tumors localized to eloquent regions or deep in the skull base may be more amenable to SRS with a fractionated schedule given the need to avoid large doses to critical structures [25]. As the size of the tumor volume at SRS is associated with survival outcomes, it is important to consider that size may have historically limited delivery of appropriate radiation doses to larger tumor volumes [10]. A wide range of volumetric cutoffs have been proposed from < 14cm3 to 24cm3 [26, 28]. Vordermark et al. reported that larger prescribed doses (> 15 Gy) up to 30 Gy were associated with improved survival [29]. Reynaud et al. escalated the dose to a mean of 35 Gy demonstrating a survival benefit [30]. In contrast Fogh et al. reported increased radiation toxicity with doses over 40 Gy which may act as an upper bound for the therapeutic window of radiation [25]. Furthermore, in the context of a propensity for local reccurence, investigation of the parameters applied to the tumor margin has been conducted. Guan et al. argued that the improved tumor control rate in their cohort was partially a result of delivering radiation to the margins with an isodose line of 63–75% resulting in higher overall marginal irradiation [26]. Consequently, debate over how to best define the clinical target volume (CTV) exists, and is made difficult in the recurrent setting as radiographic appearance of the tumor may be impacted by surgical/radiation effects the primary tumor treatment [31]. Bell et al. suggested that poor delineation of the clinical target volume was responsible for underdosing of the recurrent target volume [31]. In their work the authors demonstrated the role of the emerging whole brain spectroscopy technology [31]. Through the use of a Choline: N-acetylaspartate ratio > 2 representing disease activity, a clinical tumor volume of 2 cm resulted in total disease coverage approaching 99%, while no CTV expansion resulted in only 54% disease coverage [31]. Indeed, the authors of this study highlight the need to properly account for microscopic disease in the recurrent setting through the use of CTV expansion.

Chemotherapy has been a mainstay in the management of rHGG patients yet is limited by heterogeneity in reported clinical outcomes. While TMZ has historically been a backbone of systemic therapy, novel agents directed at more specific glioma properties have been developed [10]. A contemporary therapy under intense investigation is the anti-vascular endothelial growth factor (VEGF) monoclonal antibody Bevacizumab (BVZ). Previous studies have demonstrated that BVZ in combination with SRS potentially limits radiation toxicity, namely radiation necrosis [4, 22]. However, the role of BVZ in combination with SRS is variably associated with survival outcomes with many studies reporting no significant improvement in overall survival [4]. While more work is to be done to validate the role of BVZ in improving survival for rHGG patients undergoing SRS, there may be a role in radiation associated toxicity reduction. A methodological limitation of the current literature reporting on outcomes in patients receiving multiple lines of chemotherapy, particularly investigational therapies is a trend toward selection bias. Many of the patients in these studies demonstrate higher performance status, which may confound outcomes [29].

Reccurence proximal to the treatment volume is a central issue in both primary and rHGG. Most patients in both disease settings experience recurrent disease within a few millimeters of the treatment volume [3]. This pattern of failure is referred to as local progression [14]. A primary barrier to successful repeat resection in rHGG is the diffuse nature of growth resulting in poorly defined tumor margins [30, 31]. In part, the inability to clearly define resectable tumor may result in residual tumor growth adjacent to the resection cavity. Similarly, radiation planning may be limited in the same way when attempting to design a conformable plan to ambiguous tumor margins. Furthermore, as SRS is defined by a steep dose gradient, a portion of the dose may fail to capture marginally invading tumor cells [30,31,32]. One solution has been the introduction of extended field radiation [32]. Koga et al. demonstrated improved local control using an extended field radiation approach over conventional SRS, albeit without a significant survival benefit [33]. This approach may be promising however, larger lesions (> 20 mm) may not be eligible because an expanded radition field of such magnitude may place the adjacent uninvolved parenchyma at excessive toxicity risk [33]. However, this study did not report a significant increase in radiation toxicity in the extended field SRS cohort compared to the conventional SRS cohort [33]. Overall, the use of a so-called clinical target volume which accounts for additional treatment area outside of the enhancing lesion has been inconsistently utilized in this setting and warrants further exploration [5].

The typical adverse (AEs) noted in SRS treatment of rHGG are low grade (Grade 1) non-specific acute toxicity such as: alopecia, skin erythema, headaches, and nausea/vomiting [34]. Multiple radiation treatments over a lifespan elevate the risk of radiation induced toxicity, tissue damage, or secondary malignancy [22]. Given this, there ought to be attention to the interval between the radiation administered for the primary tumor and the recurrent tumor. In fact, Imber et al. demonstrated in their cohort of rHGGs an elevated risk of SRS related AEs in patients with a shorter interval from primary radiation treatment [22]. Additionally, the authors of this study demonstrated higher rates of radiation necrosis with treatment of larger tumor volumes [22]. These findings may be related to the associated dose escalation with large tumor volumes [22]. Large tumor volumes requiring increased dose of radiation have increasingly benefited from the adoption of hypofractionated shchemes which allow for similar biologically equivalent dose application as single session schemes with reduced adverse radiation events [25, 26]. It is also important to discriminate clinically relevant radiation necrosis from other instances. Most importantly, cases of radiation necrosis necessitating surgical intervention represent the most clinically significant cases and should be reported separately in future work. Finally, the assessment of radiation necrosis is evolving and has increasingly incorporated new diagnostic modalities like magnetic resonance spectroscopy (MRS) and single photon emission computed tomography (SPECT) [34]. Perhaps the evolution of these technologies will allow for a more accurate delineation of RN allowing for the development of consistent assement and reporting standards Alternatively, the authors of this study demonstrated an inverse relationship between radiation necrosis and time between primary irradiation and SRS [22]. Cumulative radiation exposure is a recognized risk factor for toxicity, and a patient’s history of exposure ought to be accounted for when selecting for further radiotherapeutic treatment [22]. Ultimately these findings provide support for the argument favoring fSRS in larger tumor volumes as discussed previously. There may be additional benefit from redefining how radiation necrosis in these cohorts is presented. Not all cases of radiation necrosis are clinically equivalent, and the most significant cases are those requiring additional surgery. Therefore, modifying radiation necrosis with the need for surgical intervention could contribute to more meaningful data reporting [10].

As with any systematic review the strength of the analysis reflects the individual studies included in the analysis. One of the persistent limiting factors present in most of the studies included in our analysis was the heterogeneity with respect to WHO grade. Many of these studies included both grade III and IV tumors. Higher grade tumors have consistently been associated with diminished overall survival. Given this it is important to understand that the overall survival point estimate may be an overestimate. Furthermore, our review includes patients spanning multiple iterations of WHO grading which means there are potential misclassifications of tumors in older studies relative to more recently published studies. For example, the WHO update in 2016 and 2021 may not be reflected in articles published prior to these years. Similarly, our analysis includes studies published prior to 2005 and the widespread adoption of the Stupp protocol for the management of primary Glioblastoma [36]. Studies published prior to 2005 in our analysis on average may have relatively less overall survival. Furthermore, many of the papers included in the analysis reported varying degrees of primary tumor resection. Additionally, a clear definition of gross total resection and subtotal resection was not consistently reported. As a result, this may have contributed to wide range of SRS treatment volumes, and differential growth patterns introducing excess variability across SRS planning [23]. Furthermore, the classification of “recurrent disease” is dubious, while standardized reporting of reccurence using the RANO-glioma criteria has increased in utilization, some of the earlier papers in this review did not incorporate this criteria. Another limitation included inconsistent reporting of molecular markers associated with survival including but not limited to: MGMT methylation, IDH mutation, CDK2NA homologous deletion, 1pq19 deletion, TERT and EGFR promoter mutations.

Finally, there are a number of key limitations which ought to be considered for future work. Most importantly there is a pervasive issue in the current literature as it pertains to consistent reporting of data. This is highlighted best by the absence of specific surgical as well as chemotherapy deatils across many of the studies summarized herein. Chemotherapy undoubtedly plays an important role in the recurrent setting. While the exact details of which chemotherapy regimens are most efficacious it is important that any study reporting on outcomes in this patient population make an effort to include granular details on the exact regimens employed by the study investigators. Furthermore, we believe it is imperative to define the interval between the completion of adjuvant treatment for the primary tumor to the time of first recurrence as this represents a treatment free interval and is most meaningful to patients. Many studies analyzed time from surgery to recurrence, or time from initial diagnosis to recurrence, which are important parameters but may lack relevance to patients. Moreover, some tumors may display aggressive biology, relating to a “rapidly progressive” phenotype. Consistently defining an interval by which studies can identify such characteristics may contribute to individualized follow-up and management.

To date the management of rHGG is largely individualized and lacks widespread consensus. Consequently, there is a need to further define outcomes for a variety of treatment modalities in this setting. Ultimately, support for a standard treatment protocol should be gained through prospective comparative analysis. Examples of highly impactful studies would include single fraction SRS versus fractionated SRS, and SRS with and without concurrent therapy and resection with and without SRS to the post-operative cavity to name a few. Furthermore, tumor treating fields (TTF) is an emerging technology in this space and has been previously been studied in the rHGG setting [35,36,37,38,39,40,41,42]. When combined with chemotherapy, TTF has demonstrated prolonged survival with minimal toxicity, however, there are substantial concerns regarding patient compliance [35,36,37,38,39,40,41,42]. It is imperative to limit treatment related toxicity in this patient population, and therefore TTF represents a viable alternative to SRS and therefore the literature would benefit from direct comparative analysis. To date a trend to towards success with SRS in rHGG with young patients, good performance status, and small tumor size has been reported [42,43,44,45,46,47,48,49,50]. Based on these findings, a simple nomogram incorporating these factors could prove meaningful in supporting the decision to proceed with SRS over alternative therapies. It is clear that WHO grade mediates survival outcomes; therefore, analyzing outcomes for grade III and IV tumors separately may be appropriate. Given the recent WHO update in 2021 [51], with the recognition of isocitrate dehydrogenase (IDH) mutant status, examination of IDH-wildtype and IDH-mutant pathology treated with SRS in the recurrent setting will reveal potential differences in radio responsiveness of these respective subtypes. To our knowledge this has yet to be examined in a comparative study, however as molecular signatures have been established as critical prognostic factors this may have introduced bias in studies reporting prior to 2021.

While local control is a primary endpoint for this population, distant recurence can complicate management even further. Other than Kite et al., there were no studies conducting a regression analysis of local tumor control following SRS. Further there were no studies reporting a regression analysis for distant tumor control following SRS. Indeed, in their analysis Kite et al. demonstrated a non-significant relationship between BVZ at the time of SRS on local tumor control (HR: 0.90 p = 0.81) or any chemotherapy following SRS (HR:0.71, p = 0.50 [52]. This analysis should be extended to distant tumor control as chemotherapy is likely the most influential in this aspect of disease control. Understanding the influence of a multifocal disease process on SRS candidacy deserves further investigation. Finally, an investigation of molecular signatures associated with improved radiotherapy response may better inform the decision to proceed with SRS in select cases. Previously, MGMT promoter methylation was demonstrated to portend improved OS, which may influence the selection of chemotherapy at the time of SRS [52].